Remove _call_accelerator_hook Trainer method

[daniellepintz]

What does this PR do?

After #10890 we only have one remaining usage of _call_accelerator_hook. We can remove this call by moving the line torch.cuda.empty_cache() from on_train_start to setup in accelerators/gpu.py. Then we can remove the _call_accelerator_hook Trainer method entirely.

fix #10905

Does your PR introduce any breaking changes? If yes, please list them.

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

Make sure you had fun coding 🙃

[tchaton]

LGTM !

[daniellepintz]

@four4fish why is this a breaking change? It doesn't affect users

[four4fish]

@four4fish why is this a breaking change? It doesn't affect users

for customized accelerator user who may override this function

[ananthsub]

please be sure to fill out the checklist (especially filling out a changelog entry) before this is merged